Sandia Labs FY22 Laboratory Directed Research & Development Annual Report

FY22 ANNUAL REPORT

PIRAMID: PHYSICS-INFORMED, RAPID AND AUTOMATED MACHINE-LEARNING FOR COMPACT MODEL DEVELOPMENT. The PIRAMID LDRD project advanced the possible

physics conservation laws in an ML architecture. The team’s approach extends to other domains and will be explored through collaboration with Nathaniel Trask’s DOE Early Career project. Treating other Radiation, Electrical, and High Energy Density Science models at Sandia similarly will allow for broader multi-scale modeling. The team collaborated with Xiaozhe Hu, James Adler, and Casey Cavanaugh at Tufts; George Slota and Chris Brissette at RPI; and Jack Garbus at Brandeis. An article on this project’s research was published in the Journal of Computational Physics . (PI: Andy Huang)

realization of multi-scale modeling between Radiation Analysis, Modeling and Simulation for Electrical Systems (RAMSES) simulation codes Charon and Xyce. The project team developed a workflow to train a novel ML architecture on continuum semiconductor physics simulation data from Charon for incorporation within a circuit simulation in Xyce. Compact models typically require manual calibration which might demand Charon analysis. The team workflow enables analysts to immediately leverage Charon data to produce models for use within circuit analyses. The enabling mechanism is incorporation of

HIGH FIELD AND FREQUENCY ELECTROMAGNETIC (HIFI-FREE) SHIELDING. The goals of the HiFi-Free LDRD project were: (1) create a novel physics-based model of EM shielding for cylindrical symmetries, (2) develop hybrid absorbers with shielding effectiveness (SE) greater than mu-metal at reduced density/ thickness from 100 MHz–10 GHz, and (3) absorptive from 1–10 GHz. Four hybrid absorbers were fabricated with one third the density of mu-metal and created a 0.34 mm shield that absorbed 60%–80% of incident power from 1.5–10 GHz. The team fabricated 100 nm thick Cu and Mn 0.5 Zn 0.5 Fe 2O4 multilayers and found one bilayer provided 14–18 dB of SE from 1.5–10 GHz, with negligible improvement from additional layers. A novel physics-based model of EM shielding was

investigate multilayer thin film shielding. The team, who collaborated with faculty at Penn State on some aspects of the project, characterized the absorption, reflection, shielding effectiveness and complex permittivity/permeability of 35 µm Mn1-xZnx ferrite, in steps of x=0.1, and discovered the x=0.4-0.7 stoichiometries to be the most

developed and tested that provides accurate results for dielectric and magnetic shielding from 100 MHz–1 GHz. (PI: Andrew Scott Padgett)

Absorption coefficients of SNL4 and TPL Epoxy.

Shielding effectiveness of one, five, and ten MZFO/Cu bilayers on 0.54 mm of alumina.

84